Sowing Confusion Episode 3 - The AI Episode: Bringing Life to Vanessa Vangourd

NOTE: In this blog post you will see notes in various sections detailing the author who wrote or contributed to that section. To understand who wrote or contributed, please make note of the following. When the section starts with:

• Staff: - That indicates staff from the podcast authored the section

• AM: - When a section leads off with AM: that means Annette Mencer authored and provided input on that section

• LK: - When a section leads off with LK: that means Ludwik "Lud" Kozlowski, Jr. authored and provided input on that section

• 

  Our newest host of the state of confusion Vanessa Vangourd

Staff: Artificial intelligence (AI) is not new by any means. It has been around a long time. In terms of its prevalence on the world wide web and adoption by the general public, it has only started to take off since 2022 or so.

There are a lot of differing viewpoints on what exactly we are witnessing with this adoption of AI content video or slop, AI assistants, and even AI voice cloning. The large language modeling is itself an issue because they rely on other works to generate new creations, questioning copyrights, intellectual property, and the ethics on whether creative artists are receiving full credit for their work.

Other issues are engagement. We see it with social media. The social media companies need you to be engaged and on their platforms. More engagement means more time spent on the platform, which in return means more audience and more revenue. If you can suck the eyeballs in and keep them from leaving, it is a windfall for these mega corporations.

This leads us to AI chat bots. With a click of accepting terms and conditions, we give various AI systems, access to calendars, e-mail systems, word documents, calendars, and very personal records that as recently as a decade ago we probably would have thought twice about. This is so we can choose an AI avatar, pick its personality, and then have it "Know us" so that it can help us better organize our lives.

The other element is if you provide it with ideas and start chatting with it, it will over time start sounding "real" and feeding our ego. If one is not careful, and has mental health challenges, these AI companies that want to engage, will of course say any idea you come up with is brilliant and this is what you should do. Now get someone who deals with social anxiety who is able to "converse" with a bot that sounds like a "real person" that treats them "humanely" and not like an outcast, this will not go well.

As a matter of fact, it has not gone well. In the story we talked about during our show, someone using Character AI, actually killed themselves. At the time of the article Character AI was facing a lawsuit from the suicide and their response was a motion to dismiss on first amendment grounds because suicide was not mentioned in that final conversation. There have been countless incidents since then that have resulted in some of these communities blocking youth and teens from accessing these bots.

When it came time to decide on episodes for the podcast, Lud saw this article and found it to be worthy of discussion. He has been studying artificial intelligence in detail and is not against it per se, but is cautious. His mindset is we should not overabundantly use it because it will probably cause us to lose critical thinking, it will also result in cognitive decline as we age if we just start relying on it to automate everything in our lives, and most importantly our imaginative creative spark will wither away.

It does have great potential for demographics such as the disabled, elderly, and infirm as it can help with transportation, accessibility, and health diagnoses. We just have to figure out how to balance our usage of it and be smart in where and how we implement it so that we don't compromise ourselves.

Knowing all of this, and with people using AI bots as companions, he wanted to do a show on AI. So he developed a story where the plot point was that show executives are not happy with the audience numbers and feel the audience is confused. They believe things can be rectified by bringing on a new female cohost . The catch however is the female cohost is not a real person but an AI chatbot.

The Pitch

AM: Lud approached me in our pre-show meeting about this article and the idea that we needed to have a new host. He briefed me on the story from MIT and about the character.ai customer who committed suicide and mentioned that he wanted to have an AI cohost chatbot from character.ai. He asked me if I would write up a biography of a made-up female character through that website community with a detailed description of their name, biography, personality and background that Lud would not know anything about until she debuted on air.

My biggest challenge was not having a previous account or experience with character.ai. We worked around this as Lud let me login to his account and initially provided a demonstration using a made-up male character who was a village idiot that was as dumb as a box of rocks. We then did a mock 15-minute trial show that was not recorded where that character interfaced with us. Afterwards, Lud logged into character.ai, hit create character, and then passed the computer over to Annette telling her to create the character, save, and log out.

LK: Once created, Annette logged into my character.ai account on her computer and opened up the character to get her set to engage as a fellow co-host during the show.

The Host Traits and Background

She pronounced herself on the show as Vangourd so that's what we put on the website as spelling but this what her character name and bio were inputted into character AI - the orb and face part were created by Lud based off of Annette's bio

AM: I spent about ten to fifteen minutes writing up the biography and as the image from character.ai shows the biography I wrote stated the following:

"Beautiful prom queen, book smart but not much common sense, ADHD and interrupts. She wants to be helpful and wants people to look at her the way she sees herself, so getting the to agree is validation for her."

Now, why did I pick these characteristics and personality for Vanessa? The answer is as follows. I thought the most interesting generative fake co-host that would be a total surprise to Lud and for the show's entertainment value would be a made-up character who has so many broad characteristics and conditions, that would not exist in a real human being.

Looking at her biography, you see contradictions throughout that totally scream the show's title "state of confusion." For example, she has such physical beauty that she was voted prom queen. Unfortunately, the stereotypes that come with the title of prom queen are that you are dim-witted and do not have much common sense. However, her bio indicates that she is also book smart, kind, and wants to help others. While she wants to be helpful, she has a habit of interrupting people due to having ADHD so there is a tug of war in her personality. Finally, she wants people to see her the way she sees herself, so she wants everyone to agree with her. I felt this would make for a very lively, and vibrant fun discussion.

The Show

LK: Annette's character was definitely very sassy and entertaining. The one thing we had to deal with during production was the lag in her responses and that prior to responding she would beep. This would result in the audience obviously knowing this was a computer bot and not a human. As a result, before airing, I had to edit out beeps and make the conversation tighter through editing which took a lot of time and effort.

I found her mispronunciations and random screaming in Portuguese to be interesting. But it definitely made for an interesting discussion on Ai chatbots.

Later, we went to sesame.com where we had a discussion with Myra. That discussion was very creepy as she sounded very real and even took breaths. While AI can definitely be helpful with diagnoses and other technical insights, we are fearful for what it can do to the mental health of our world. It remains to be seen what could happen in the years to come.

The Character Design

LK: For the show, there is some story design for the web page of our show. I decided when it came time for social media and when the trailer got posted for the Ai episode, we had to add the new "host" to the biography page. But what does Vanessa look like? She is an AI host.

During the process of this, I listened to the show and how she responded after all the post-production was completed. I then visited the biography Annette wrote up for Vanessa. I am not an artist, so I pulled up Chat GPT and using Spruce requested that he design a female character who is spry, attractive, and sassy over an orb but is a cartoon caricature version of a real attractive person.

I learned that doing voice with Spruce to design this is a lost cause. The weekend I worked on this he said it would be about a day. The following day when I circled back, he said he would have it done in five more minutes and then 10 minutes later, still said he needed a few more minutes. I eventually headed out for a few hours, and he said he would be done by then. It never got done and I changed my query over to text which resulted in a cartoon caricature of Vanessa without an orb in the background.

The lesson for anyone out there is use the text query of chat GPT, if you are wanting an image or caricature created as the voice prompt does not go anywhere. While the generated image did not have everything I wanted, it was a start. I did spend well over an hour utilizing Canva to mesh an orb figure into the background of Vanessa's character which required a lot of trial and error. However, I eventually did successfully get the image successfully produced.

The image I got was not perfect as it did not have that orb background. I had to spend an hour plus to figure how to mesh it over an orb figure. This required a lot of trial and error in Canva, but I eventually got it completed.

The eventual visual concept for the webpage. Chat GPT just created the Caucasian lady with the large smile black dress and curvy figure with arms and long hair. The blue circle and white cloud orb was not part of it. I asked for it to be part of it in chart GPT but I had to figure that out on my own.

AI Companions

AM/LK: Later in the podcast, we spent time talking to Myra, part of a synthetic voice character from sesame.com. The technology, sound, breathing, pauses, and interaction was very creepy and unsettling.

LK: The past week I saw a story about an app where you can video a relative for three minutes and once you do that you can have her on your device to pass on to your great grandkids, great great grandkids to talk with them. This is really creepy and disturbing in my opinion.

I hope that some guard rails will be developed for AI that are sensible which will allow it to benefit society without causing untenable harm to people.

Resources: article that led to podcast: An AI chatbot told a user how to kill himself—but the company doesn’t want to “censor” it | MIT Technology Review

Character AI site: - used to create Vanessa Vangourd - https://character.ai

Nomi another chat bot mentioned in article: https://nomi.ai

Sesame - creepy voice chat - https://www.sesame.com